Subquadratic

SubQ Snap is a long-context decision model from Subquadratic, built for enterprise work where a system has to read a lot of material and return a decision fast. Subquadratic says Snap is the only decision model or LLM where compute per token stays constant as context grows. It accepts 384K tokens of multimodal context, which the company says is 12x larger than Jev, and scores 61.7 on Decision Index v0.2.1 versus Jev's 57.9, outperforming on 13 of 16 enterprise benchmarks. Speed and price are the headline. Snap processes 100K tokens and returns a decision in under one second, with stated model speeds of 80ms at minimal context, 250ms at 32K, 660ms at 100K and 2.5 seconds at 384K. Pricing is $0.05 per million input tokens, and output is free. The model is available today to anyone, and the first 10,000 sign-ups get 100M free input tokens; the team notes capacity may run out quickly and latency will vary while the platform is tuned. Subquadratic previously launched SubQ, its sparse-attention long-context LLM. Co-founder Alexander Whedon (@alex_whedon) posted the SubQ Snap launch video on October 12, 2026. Access is at subq.ai.

Be the first to comment.